SANKET: Signal-to-Safety Engine

Project: https://app.melius.com/projects/d73be6ea-a114-4b90-8dc5-79b3cc55053c/canvas/b10709a6-a2f3-4d4d-af17-329c390ecdb7

Video walkthrough: https://youtu.be/BqnWSIbRLo4

X post: https://x.com/manjunath_3155/status/2056991326441488778

SANKET is a public-safety media workflow built entirely in Melius. It turns one raw crowd-pressure signal into calm, multilingual, deployable public-safety communication.

The fictional scenario is Nandikeshwara Yatra, where Gate 3 becomes heavily crowded after evening aarti. Families, elderly pilgrims, vendors, and volunteers are present, and the system needs to redirect people toward Gate 5 and a shaded waiting corridor without creating panic.

Instead of producing only one poster or one video, SANKET acts like a reusable emergency communication engine. It uses Mel as a multi-role safety team: risk analyst, crowd-flow planner, language adapter, visual director, and public-trust editor.

The workflow generates: • LED-board messages • WhatsApp/SMS alerts • PA announcement scripts • volunteer instructions • safety posters • route diversion graphics • social alert cards • a stitched vertical safety video

Languages: English, Hindi, Kannada, Tamil, and Telugu.

Process:

Started with a fictional crowd-risk incident signal near Gate 3.

Defined the audience: families, elderly pilgrims, vendors, volunteers, and multilingual visitors.

Used Mel to analyze risk, crowd flow, language, visuals, and public trust.

Generated multilingual safety communication for LED boards, WhatsApp/SMS, PA systems, and volunteers.

Created a production-ready visual media kit with posters, route diversion graphics, social cards, and LED mockups.

Generated multiple safety video scenes and stitched them into one vertical guidance video.

Organized the canvas into a judge-friendly workflow from signal input to deployable public-safety media.

Node types used: custom text, Mel/agent reasoning, multilingual copy, image generation, video generation, stitched video, and final summary board.

Feedback on Melius: Melius made the whole creative process visible. Instead of a black-box AI output, I could build an inspectable workflow where strategy, copy, visuals, and video stayed connected on one canvas. Mel helped turn a vague public-safety idea into a structured multi-node creative system.

Final takeaway: Detection is only half of public safety. Once a crowd risk is detected, people need clear, trusted, culturally aware communication immediately. SANKET is the missing creative layer between a crowd-risk signal and public action.